Activity Number:
|
399
|
Type:
|
Contributed
|
Date/Time:
|
Thursday, August 15, 2002 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics & the Environment*
|
Abstract - #300866 |
Title:
|
Recommendations for Statistical Analysis of Groundwater Monitoring Data
|
Author(s):
|
Molly Leecaster*+ and Thomas Stoops and L. Koeppen
|
Affiliation(s):
|
INEEL and INEEL and INEEL
|
Address:
|
PO Box 1624, Idaho Falls, Idaho, 83415-3779, USA
|
Keywords:
|
Detection ; Shewart-CUSUM ; Control Chart ; Prediction Limits ; Groundwater Monitoring ; Radionuclide
|
Abstract:
|
The Idaho National Engineering and Environmental Laboratory has monitored the Snake River Plain aquifer underlying the Radioactive Waste Management Complex since 1971. Detection of radionuclides was based on the sample-specific minimum detectable activity and radioanalytical uncertainty. This definition of detection provides no comparison with background levels. Low-level analytical detections have all been discounted on the next quarterly sample. This method is error-prone since contamination would be persistent under the assumed diffusive plume model. These false-positives cause undue public concern and political upheaval. We make two recommendations to increase credibility in our data evaluation process, adding a statistical procedure to compare sample results to background status and timely re-sampling of the aquifer. We compare three statistical procedures: a combined Shewart-CUSUM control chart and parametric and nonparametric prediction intervals, on the number of detections from historical monitoring data. The combined control chart, which is sensitive to large deviations, as well as gradual trends, and re-sampling produces the fewest false-positives.
|